Approved Cluster Applications

Farrell, Paul
Dept. of Math and Computer Science
Kent State University
Cluster Acquisition for Communications Optimization and Computational Steering

Co-PI
Arden Ruttan

Abstract: It is proposed to investigate issues related to increasing the efficiency of point-to-point and collective communication on distributed clusters consisting of SMP nodes connected by high speed networks such as Myrinet, Giganet, and Gigabit Ethernet.

In addition, it is proposed to use the equipment in the development of an environment for interacting with and steering large scale scientific computation running on clusters.

Melton, Austin
Dept. of Math and Computer Science
Kent State University
Cluster Ohio Grant Program

Co-PIs
Laura Bartolo
Paul A. Farrell
Eugene C. Gartland
Mark W. Kershner
Michael Lee
Arden Ruttan
James L. Blank
Daniela Calvetti
Arne Gericke
Robert Heath
Douglas W. Kline
Lothar Reichel
Christopher Woolverton
G.Q. Zhang

Abstract: This proposal does not represent a single research project; rather, it represents a number of research projects from faculty members in northeast Ohio. These projects range from cell biology where we are enhancing confocal microscopy images and trying to develop a software model of a completely viable prokaryote cell to investigating predator-prey interactions in large lake ecosystems to numerical modeling of optical devices based upon "liquid crystal "materials to theoretical computer science research questions. Further, we will study the managing, storing, and processing of data in the cluster environment. A strong point of this proposal is that these relatively few machines will significantly enhance the computing resources of several researchers in Northeast Ohio. Further, since KSU has a very good Systems Staff, these machines will be well maintained for use by the researchers at KSU and in NE Ohio.

Wang, Bin
Dept. of Computer Science and Engineering
Wright State University
Design and Implementation of a Parallel Simulation Environment for Network Research and Education

Abstract: Computer-based network simulation is a computation and memory intensive task. For next generation internetworks, network configurations of extremely large size have to be simulated to study issues such as scalability, survivability, and/or packet loss correlation in arbitrary topological, multi-domain networks. Therefore simulations may require models to contain millions of network nodes. It has been recognized that the capabilities of conventional sequential network simulation techniques are inadequate to address such simulation requirements, and that parallel simulation techniques must be brought to bear on these challenges.

However, {\em parallel network simulation} has not yet been widely embraced by the network research community due to a number of difficulties: (1) Lack of established, easy to use modeling methodologies suitable for parallel execution; (2) Absence of mature software environments and comprehensive application demonstrations. The objective of this project is thus to {\em propose a component-based parallel network simulation software architecture; develop, and demonstrate, an efficient (in terms of simulation speedup) and user friendly parallel network simulation environment for the state-of-the-art network research and higher education}. The parallel simulation run time infrastructure developed in this project will intelligently incorporates the idiosyncrasies of networks and can thus very well take advantage of the computational power and physical memory offered by PC clusters.

The PC cluster based parallel simulation environment developed and the technologies demonstrated in this project will have a positive impact on networking research nationwide by providing a more efficient and user friendly parallel simulation tool. It will also provide a good teaching tool to be used by undergraduate and graduate students at Wright State University and other institutions in their introductory and advanced communication network courses for the education of a generation of information technology work-force. We intend to make the technology and software developed in the project accessible to OSC, industrial organizations, and other academic institutions. The PI will cooperate with researchers at OSC on the transfer and redistribution of technology and software developed in this project without restriction. The PI and the department of computer science and engineering at Wright State University are also committed to make any idle cycles available to the Ohio research community.

Gallagher, John C
Dept. of Computer Science and Engineering
Wright State University
Cluster Computing for Bioinformatics and Biocontrol

Abstract: Recent advances in both human biology and micro-fabricated devices provide knowledge that will allow the development of technologies that directly and positively impact human health and quality of life. However, the sheer volume of the data involved as well as the combinatorics of evaluating potential engineered devices represents a huge hurdle to both scientific analysis and engineering progress. Parallel computation, properly applied, can help us to overcome these hurdles and develop meaningful, life-improving, devices and techniques more quickly than might otherwise be possible. This proposal will detail three computational biology projects that will have positive impact on public health and that will be enabled by parallel computation and OSC support through the Cluster Ohio Program.

Amar, Jacques
Dept. of Physics and Astronomy
University of Toledo
Acquisition of a Beowulf Cluster for Research & Development of Advanced Computational Physics Modeling

Abstract: The acquisition of a 4-node, 16 processor Beowulf cluster along with a 2-processor server is proposed to develop and carry out large-scale parallel computations in condensed matter physics, surface physics, and astrophysics. We will carry out investigations in three project areas: simulations of epitaxial thin-film growth and nanomachining (Amar), electrical discharges in plasma displays (Theodosiou), and hydrodynamics of and radiation transfer in astrophysical outflows (Bjorkman). Two of these projects involve processes of great current technological interest, while the other involves fundamental investigations of astrophysical environments.

Since these projects are amenable to parallel computing by using domain decomposition, a Beowulf cluster appears ideally suited. The proposed projects are expected to lead to significant advances in our ability to simulate extended time and length scales as well as significant advances in our understanding of thin-film growth, plasma discharges, and astrophysical outflows. The proposed cluster will also be used as an experimental facility to develop improved parallel algorithms, investigate alternate network topologies, and implement various strategies for parallel computation. As such, it will also serve as a seed for the eventual development of a larger parallel facility at the University of Toledo. In addition, it will serve as an important resource for undergraduate interest, such as parallel software, networking solutions, and communications strategies, we will generate mutually beneficial tools, leading to major advances in out computational physics programs.

Crescimanno, Michael
Dept. of Physics and Astronomy
Youngstown State University

Cluster Ohio Grant: The YSU Advanced Computation Working Group Cluster

Co-PIs
Robert Kramer
Nathan Ritchey
Mark Welton
John Buoni

Abstract: We propose locating a cluster of 32 processors at YSU as part of the Cluster Ohio grant. In brief, the cluster nodes could be upgraded with faster NIC cards and tied together with a pair of Giga-Ethernet switches; these items will be purchased through a combination of external moneys and internal YSU moneys. The cluster will facilitate the efforts of the Advanced Computation Working Group at YSU. The Working Group staff are actively seeking outside financial support for the long term (in the form of Research Professorships) and some external moneys for release time to set up and to run their research code on the machine in the coming academic year. The science and educational scope of the project includes computational chemistry, studies in parallel algorithm development, mathematical biology, theoretical atomic physics, teaching students and staff how to program in a massive parallel environment, and teaching students the practical arts of machine administration and maintenance.

Elster, Charlotte
Dept. of Physics and Astronomy
Ohio University
Modeling Complex Systems on a Beowulf Cluster

Co-PIs
Peter Jung
Paul Gailey
David Drabold
Thomas Statler

Abstract: A 24-CPU Intel cluster is requested to provide local computational support for projects that model dynamics of nuclear systems, analyze complex dynamics in biological systems, carry out ab initio molecular dynamics simulations of amorphous semiconductors and glasses, and describe eccentric structures in galactic nuclei. This combined request for a larger system will optimize the benefits of a locally available cluster and create a better staging environment for the Pi's to be more competitive within their individual disciplines in applying for larger resources.

White, Edward
Dept. of Mechanical and Aerospace Engineering
Case Western Reserve University
High Performance Computing for Computational Mechanics

Abstract: Research in the Case Western Reserve University Department of Mechanical and Aerospace Engineering ranges across a number of areas of mechanics: fluid mechanics, solid mechanics, and biomechanics. All of these areas are being transformed by rapidly expanding computational capabilities especially through the development of parallel computational approaches. The local computational resources in the department presently consist of a variety of single and dual-processor workstations. This limits in-house computational research to fairly small-scale problems and drives more ambitious research to outside agencies. The proposed 8-machine, 32-processor parallel computing cluster would greatly enhance the computational capabilities of the department and would permit much more ambitious computational problems to be undertaken locally. A department-wide parallel computing resource would have a number of important benefits including greater collaborative opportunities both within the department’s faculty and with outside researchers, introduction of modern parallel computing methods to graduate students, and the ability to undertake much more competitive and ambitious research.

Liszka, Kathy
Dept. of Mathematical Sciences
University of Akron
Beowulf Cluster for Multidisciplinary Research in a Parallel Computing Environment

Co-PIs
Jianping Zhu
Ping Yi
Yingcai Xiao
C.C. Chan
Zhong-Hui Duan
Ben Yu-Kuang Hu
Edward Lim
Jutta Luettmer-Strathmann
David Modarelli

Abstract: The work proposed here addresses three major aspects of large-scale computing in a parallel environment: the investigation of specific complex systems, the development of parallelization strategies and algorithms that take advantage of cluster-style parallel computing, and the development of tools for computation and system management in a distributed computing environment. The project team consists of ten faculty with diverse backgrounds in the mathematical, computational, and natural sciences. The cluster will initially serve as the focal point of a local community of users of parallel computing environments. With initial successes, we hope to encourage more interdisciplinary collaborations both within the University and among other researchers at neighboring colleges and Universities. Current parallel programming courses offered will incorporate the cluster for program development. We anticipate that other scientific and numerical courses will benefit from cluster use as well.